Wearables Wednesday: The Need for Individualized Mental Health Models

I often think about two seemingly contrasting ideas that inform how wearables are used to monitor and improve mental health ⌚

1. The most accurate way to know about someone’s mental health is to ask them.

2. Humans are pretty terrible at accurately recalling their prior moods, thoughts, and experiences, or understanding how they have changed over time.

The first idea is why you should never rely solely on physiological data to tell how someone feels. There is a reason why most emotion-detecting algorithms are compared against self-reported measures of mood as the ground truth: you can’t get more accurate about how someone feels than by asking them.

At the same time, the second idea is why we shouldn't only rely on someone's perceptions and memory to tell us how they felt last week, last month, a year ago, etc. and whether that has changed over time, if their behavior (e.g., health behaviors like exercising, or social behaviors like meeting up with friends) is influencing their emotions, and so on and so on.

But where human limits exist, wearables can assist.

This is why I love using wearables to help assess and monitor mental health and especially stress. Wearables can help take the guesswork out of what happened when, how things have changed, and how emotions and physiology are correlated with behavior. But what we’ve seen is that these models don’t function very well on their own. Physiology is too different person-to-person and in different contexts. To be accurate and effective, they need some kind of user input.

That’s why I believe that you must allow for self-reporting when using wearables to assess and track mental health. Not only do self-reported data help give a bigger picture of someone’s health, but the algorithms can learn from the individual by incorporating self-reported data into the calculations and recommendations. That doesn’t mean you have to include long surveys—just smart items, pinged at the right times.

To me, the dreamiest wearable would track my physiological and behavioral signals (including location variance—a highly valuable but overlooked marker of depression) and periodically request self-reported data to train the model on what wearable-derived metrics are more meaningful for my mental health. The result would be an ultra-individualized wearable that learns from the individual by incorporating self-reported data into the calculations, with AI-assisted recommendations.

What do you think? Can wearable-derived mental health metrics exist without self-reported data? If not, where does the balance lie? 🤔

---

👋 I'm Lydia Roos, health psychologist and Founder of EvolveWell Research Partners.
📌 EvolveWell offers consulting and research services for health, wellness, and fitness companies. [evolvewellresearch.com]
📩 Get in touch via email or LinkedIn.